SiFive raises $400M
Chip designer SiFive closed a $400 million funding round led by Atreides to scale its business into AI data‑center markets and expand silicon offerings. The financing underscores investor appetite for RISC‑V based designs as alternatives to established chip architectures. (x.com)
SiFive just pulled in $400 million at a $3.65 billion valuation, and the money is aimed at one of the hardest places to break into in chips: the data center, where companies buy processors by the rack and expect them to run artificial intelligence workloads around the clock. (sifive.com) The round closed on April 9, 2026, and SiFive said it was oversubscribed, which means investors offered more money than the company chose to take. Atreides Management led the financing, with Nvidia, Apollo Global Management, Point72 Turion, T. Rowe Price Investment Management, Prosperity7 Ventures, and Sutter Hill Ventures also participating. (sifive.com) SiFive does not manufacture chips in giant factories like Intel or Taiwan Semiconductor Manufacturing Company. It sells processor designs and related intellectual property, which other companies license and then build into their own chips. (reuters.com) The bet here is on RISC-V, pronounced “risk five,” which is the basic instruction language a processor understands. RISC-V is an open standard, so companies can build on it without being locked into a single owner’s architecture. (riscv.org) That makes RISC-V the chip world’s version of a public road instead of a private toll road. By contrast, Arm licenses a proprietary instruction set, and Intel’s x86 architecture has long dominated servers and personal computers. (riscv.org) SiFive has been tied to RISC-V from the start because it was founded in 2015 by Krste Asanović, Yunsup Lee, and Andrew Waterman, who were among the people behind the architecture at the University of California, Berkeley. The same year, RISC-V International was set up to guide the standard’s ecosystem. (sifive.com) For years, most RISC-V chips showed up in smaller jobs like embedded systems, where a processor might run a storage controller, a sensor hub, or a simple device inside a larger machine. The new push is to move that open-standard approach into servers that handle web services, storage, video streaming, and artificial intelligence inference. (sifive.com) SiFive started laying that groundwork in August 2024 with the Performance P870-D, a processor design it said was targeted at data center workloads with lots of parallel tasks. The company says that design can also pair with its Intelligence family, which is its line of artificial intelligence acceleration technology. (sifive.com) It added another piece in September 2024 with the Intelligence XM Series, which introduced a scalable matrix engine for artificial intelligence work. In plain terms, that is hardware built to do the dense, repetitive math that large models need much faster than a general-purpose processor can do it alone. (sifive.com) The investor list says almost as much as the funding total. Nvidia joined the round even though SiFive is trying to supply more of the central processor side of artificial intelligence systems, which suggests big chip companies want exposure to a future where RISC-V becomes part of the standard data center stack. (reuters.com) This is also a very different position from 2021, when Intel explored buying SiFive for more than $2 billion and the talks ended without a deal. Five years later, SiFive is still independent, newly valued at $3.65 billion, and raising one of the largest rounds yet around the idea that open-standard processor designs can win real estate in artificial intelligence servers. (reuters.com, sifive.com)
